A QOS with this flag set will not have its GrpTRESMins,
GrpWall or UsageRaw decayed by the global PriorityDecayHalfLife or
PriorityUsageResetPeriod values.  Such a QOS will act as a time-limited
quota of resources.

This flag only affects the QOS and not the associations that run in the
QOS.  Those usage values will continue to be decayed.

The only way to reset this value is with

sacctmgr mod QOS $qos set rawusage=0

If you remove the flag usage will begin to be decayed as normal.

For job resize, correct logic to build "resize" script with new values.
    Previously the scripts were based upon the original job size.
bug 3498

Do not enable init scripts if not present.

Please note that, unlike the init scripts, service files are not
automatically enabled at this time.

Bug 3371.

If node boot in progress when slurmctld daemon is restarted, then allow
    sufficient time for reboot to complete and not prematurely DOWN the node as
    "Not responding".
bug 3494

Add new advanced reservation flag of "flex", which permits jobs requesting
    the reservation to begin prior to the reservation's start time and use
    resources inside or outside of the reservation. A typical use case is to
    prevent jobs not explicitlg requesting the reservation from using those
    reserved resources rather than forcing jobs requesting the reservation to
    use those resources in the time frame reserved.

Add new advanced reservation flags of "weekday" (repeat on each weekday;
    Monday through Friday) and "weekend" (repeat on each weekend day; Saturday
    and Sunday).

Could result in squeue abort
Coverity error CID 44969

Reduces possibility of old data if job_id or user_id option specified
  with iterate option
Coverity error CID 44783

Increased maximum file size supported by sbcast from 2 GB (32-bit integer
    to 64-bits). This required changing the file broadcast RPC and several
    internal variables.
bug 3485

by specifying "--uid=<uid>|-u <uid>".

# Conflicts:
#	NEWS

A 17.02 controller,sacctmgr couldn't talk to a "master/17.11" DBD
because the 17.02 client was talking attempting to talk to the DBD with
the 17.02's MIN_PROTOCOL_VERSION -- which was 15.08 and is more than 2
version behind the master. The master's MIN_PROTOCOL_VERSION is 16.05,
so it couldn't unpack the messages.

The controller should always communicate at it's current protocol to the
DBD.

For federations, it's possible that a higher version controller could
talk to a lower version controller. So the cluster needs to talk to the
remote cluster using the remote cluster's protocol version -- which is
given back from the DBD.

Now that sched/wiki + sched/wiki2 are gone, almost all
of the sched plugin interface are unused. So remove them.

Three of these were encapsulating a call into the gang scheduler,
handle these directly where needed instead of through the sched plugin.

Enable through SchedulerParameters. Will sort by youngest jobs first,
rather than based on priority. Use alongside 'preempt_strict_order' if
you don't want the plugin to try to further optimize the preemption
list.

Bug 3457.

Introduced by commit 059275f6 when the timer is trigger.
Releasing the locks means that job_ptr may point to an element that was
deleted by a different thread in the meantime. Restructuring the code
to advance the iterator prevents this - the iterator itself does not have
this issue as the List structure will manage the position during the
sleep().

While here, move the reservation update handling outside of this
loop to simplify operation. This does not need to piggy-back on the
scan of the job_list - switching to using list_for_each should
mitigate some of the performance loss by needing a second full pass.

Bug 3414.

These were mis-calculated previously, and are internal implementation details
that weren't meant to be exposed.

association GrpWall limit.

limits.

Bug 3476

old ones.

This is cosmetic only, no code change.

Bug 3476

Bug 3472

regcomp() is not safe to use across a fork in older glibc versions.
Reinitialize the keyvalue_re structure after the fork through an atfork()
handler.

Bug 3276.

This is a regression from commit b818dd9d

Basically the first AC_LINK_IFELSE sets whatever compiler we
are using to be that.  Since the above commit removed the BGL/P
code that was linked using C C++ became the compiler since the
next thing was BGQ in configure.ac to test against.

I just grabbed the DATABASES call, but any other one could had
worked.

another one.

Bug 3465

This is the way it is done with the task plugins.  It appears this only
really matters when requesting 1 task with a full socket with exclusive
access.  This code would cyclically allocate sockets to the step instead
of filling up one socket then going to the next.

Fix for job constraint specification with counts, --ntasks-per-node value,
    and no node count.
bug 3470

Task/cray: Treat missing "mems" cgroup with "debug" messages rather than
    "error" messages. The file may be missing at step termination due to a
    change in how cgroups are released at job/step end.

Honor --ntasks-per-node and --ntasks option when used with job constraints
    that contain node counts.
bug 3458

This reverts commit 8ea967d5.

Defer interactive job allocation until ALL allocated nodes are ready
  rather than after PrologSlurmctld (if any) completes.

Bug 3467.

Add scancel "--hurry" option to avoid staging out any burst buffer data.
bug 3462

Only available in squeue with the '-O lastschedeval' format
option, no single-character option provided.

Bug 3122.

burst_buffer/cray - Do not execute "pre_run" operation until after all nodes
    are booted and ready for use.
bug 3461